The Impact of Trade Agreements: New Approach

Swarnali Ahmed Hannan
Strategy, Policy and Review Department
International Monetary Fund
Email: [email protected]
The views expressed are those of the author and should not be attributed to
the IMF, its Executive Board, or its management.
1

The Endogeneity Issue

“…by and large [the studies] fail to address the
endogeneity related to many of the policy
variables…There are many examples where the
countries that sign a trade enhancing agreement
already trade a great deal together (NAFTA, EU).”
Head and Mayer (2014, pp. 162)
The Widely Different Impact of Trade
Agreements—Baier, Yotov, and Zylkin (2016)
 Synthetic Control Method



First study to employ SCM across a large number of
trade agreements.
Current trade slowdown witnessed in data
What drives trade?
 What can policy do?

2
3





SCM is an econometric tool for comparative studies
where the control unit is determined by a systematic
data driven procedure.
SCM creates a synthetic (artificial) control unit that is
a weighted average or linear combination of the
untreated units.
The weights are chosen such that both the outcome
variable and its observable covariates/determinants
are matched with the treated unit before treatment.
The evolution of the actual outcome of the treated
unit post- treatment is then compared against the
outcome of the synthetic unit, and the difference is
interpreted as the treatment effect.
Intuitively, the SCM basically uses a weighted average
of the outcome of the control units to estimate the
counterfactual outcome of the treated unit.
4



By constructing a counterfactual, SCM can address the core
endogeneity issue related to “countries that have trade
agreements are natural trading partners and would have
traded anyway”.
Currently a very popular approach of comparative case
studies in both micro and macro studies (e.g. impact of
cigarette sales tax, economic impact of German
reunification).
Econometric benefits compared to traditional approaches:
A number of methods have been used to deal with the problem
of selection bias in observational data, including matching
estimators, difference-in-differences regressions, etc.
 These techniques are useful but do not deal with unobservable
country heterogeneity. At best, control for time-invariant
country characteristics (Hosny, 2012).
 SCM can allow the effects of unobserved confounders to vary
with time (Abadie et al., 2010).

 Coverage:
Balanced Sample
1983-1995  104
pairs
 Export – Import
 For some
exercises also
considered 19732001216 pairs
(All)
26
30
AM-AM
EM-EM
AM-EM
EM-AM
18
30
6
7
Exports, Average of
104 Country Pairs
12000
10000
12000
USD Million
10000
8000
8000
6000
6000
4000
Treated
Synthetic
2000
0
4000
2000
0
-10
-8
-6
-4
-2
0
2
4
6
8
10
The y-axis refers to ten years before and after trade agreement.
8
Exports, Average of All Country Pairs in NAFTA
140000
120000
140000
USD Million
120000
100000
100000
80000
Treated
80000
60000
Synthetic
60000
40000
40000
20000
20000
0
0
-10
-8
-6
-4
-2
0
2
4
6
8
10
The y-axis referes to ten years before and after trade ag
The y-axis refers to ten years before and after trade agreement.
9
U.S. Exports to
Canada and Mexico
350000
300000
USD Million
350000
160000
300000
140000
250000
250000
200000
200000
150000
150000
100000
100000
50000
50000
0
0
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10
Treated
Mexico Exports to
Canada and U.S.
160000
140000
USD Million
120000
120000
100000
100000
80000
80000
60000
60000
40000
40000
20000
20000
0
0
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10
Synthetic
Treated
Synthetic
The y-axis refers to ten years before and after trade agreement.
10
120
Export Growth of Average Treated Over Ten Years,
Relative to Average Synthetic (cumulative, percentage points)
100
80
60
40
20
0
EC'86
EM-AM AM-AM EM-EM
All
NAFTA AM-EM
11
Export gains over ten years (ppt)
800
700
600
500
400
300
200
100
0
-100
-200
0
0.2
0.4
0.6
0.8
1
Goodness of fit between treated and synthetic prior treatment
Size of bubles represents nominal GDP (USD million) of exporting country
during the year of trade agreement. Export gains are export growth of
treated over ten years relative to synthetic, in cumulative percentage points.
Goodness of fit is the normalized root-mean-square deviation between
treated and synthetic for the ten years prior to treatment. A smaller number
of goodness of fit indicates a better fit.
12
250
Export Growth Over Ten Years
(cumulative percentage points)
400
350
200
150
Export Growth Over Ten Years
(cumulative percentage points)
300
Trade agreements
with higher depth
250
200
100
Trade agreements
with higher depth
150
100
50
50
0
0
PTA
FTA
Customs Union
2
3
5
7
Source of trade agreements’ depth:
Left hand chart: Economic Integration Agreement Database (1950-2011), Bergstrand and Baier.
Right hand chart: Dür, Andreas, Leonardo Baccini, and Manfred Elsig. 2014. “The Design of International Trade Agreements:
Introducing a New Database.” Review of International Organizations 9(3), 353-375.
13
14

Concept:


Process:






Assess whether the effect estimated by the synthetic control for a country pair
affected by the trade agreement is large relative to the effect estimated for a
country pair chosen at random.
Randomly select 10 treated units.
Let A = exporter in the treated unit.
Randomly select 5 country pairs showing the exports of A to a country not in the
trade agreement (placebo).
Run SCM on these selected country pairs.
Compare treated relative to synthetic for treated unit versus the placebo unit.
Example:





Treated unit is CAD  USA (one of the 10 randomly chosen treated unit).
Here, CAD is the exporter in the treated unit.
Take CAD, and randomly choose 5 country pairs showing CAD exports to other
partners not in trade agreement (placebos).
Run SCM on each randomly chosen country pair.
Compare treated relative synthetic of CAD USA with that of the 5 placebo
units.
2500
Export Gains over Ten Years
(percentage points)
2000
1500
1000
500
0
-500
-1000
Placebo
Treated
16
17
-What happens to the top importer outside trade agreement?
-Apply SCM to the top importer that is outside the trade agreement.
Import Growth of Average Treated Over Ten Years,
Relative to Average Synthetic (cumulative, percentage points)
60
50
40
30
20
10
0
-10
-20
-30
-40
EM-AM
All
AM-AM
18
-What happens to the top export destination outside trade agreement?
-Apply SCM to the top export destination that is outside the trade agreement.
Export Growth of Average Treated Over Ten Years,
Relative to Average Synthetic (cumulative, percentage points)
25
20
15
10
5
0
EM-AM
All
AM-AM
19
 Trade
agreements can generate substantial
gains, particularly for emerging markets.
 The study falls under a small group of
literature that shows trade agreements
matter!
 Relevant for policy making in the current
context of trade slowdown.
 The limitations of SCM approach should also
be borne in mind while interpreting these
results.
20
Background Slides
21
There are J+1 units (regions) in periods t=1,….,T.
 Region “one” is exposed to the intervention
during periods T0+1 to T.

is the outcome that would be observed for
region i at time t in the absence of intervention.

is the outcome that would be observed for
region i at time t if region i is exposed to the
intervention in periods T0+1 to T.

is the effect of the
intervention for unit i at time t for t>T0.
 AIM: estimate the effect of the intervention on
the treated unit

22

Suppose








is given by a factor model:
is an unobserved (common) time-dependent factor,
is a vector of observed covariates
is a vector of unknown parameters
is a vector of unknown common factors
is a vector of unknown factor loadings
are unobserved transitory shocks
: heterogeneous responses to multiple unobserved
factors.
Basic idea: reweight the control group such that the
synthetic control unit matched and (some) pretreatment of the treated unit, . As a result, is
automatically matched.
23
 Let
 Each
value of W represents a particular
weighted average of control units.
 The value of the outcome variable for each
synthetic control indexed by W is:
 Suppose
 Then
that we can choose W* such that:
an unbiased estimator of
is
24
 In
practice, the vector
is optimally
chosen to minimize the following pseudodistance:
where represents a vector of preintervention characteristics of the treated
region, while is a matrix containing the
same pre-intervention variables of the
control regions.
25

Start off with the typical gravity equation used to model bilateral trade.


The dependent variables can be regarded as covariates of SCM approach.















xijt = GtMexit Mimjt φijt
Distance between the bilateral pairs
GDP of each country in the bilateral pair
GDP per capita of each country in the bilateral pair
Population of each country in the bilateral pair
Bilateral Real Exchange Rate
Remoteness of each country in the bilateral pair, proxy for multilateral trade
resistance (MTR) term (remoteness due to physical distance and/or policy).
Colonial history = 1 if pair ever in colonial relationship
Col to = 1 if export from hegemon to colony
Col from = 1 if export from colony to hegemon
Contig = 1 for contiguity
Comleg = 1 for common legal origins
Comcur = 1 for common currency
Common language = 1 for common official language
Flow, lagged by 3years
Source: Head, Mayer and Ries (2010), WDI, National Sources
26
Background Slides
27
Country Pairs
Exporting
country
1
AUS
2
NZL
3
AUT
4
BEL
5
CHE
6
DEU
7
ESP
8
ESP
9
ESP
10
ESP
11
ESP
12
ESP
13
ESP
14
ESP
15
ESP
16
ESP
17
ESP
18
FRA
19
IRL
20
ITA
21
NLD
22
NOR
23
PRT
24
SWE
25
CAN
26
USA
*Entry into force
Importing
country
NZL
AUS
ESP
ESP
ESP
ESP
AUT
BEL
CHE
DEU
FRA
IRL
ITA
NLD
NOR
PRT
SWE
ESP
ESP
ESP
ESP
ESP
ESP
ESP
USA
CAN
Year of Trade
Agreement*
1983
1983
1986
1986
1986
1986
1986
1986
1986
1986
1986
1986
1986
1986
1986
1986
1986
1986
1986
1986
1986
1986
1986
1986
1989
1989
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
Exporting
country
IDN
IDN
IDN
IDN
MYS
MYS
MYS
MYS
PHL
PHL
PHL
PHL
SGP
SGP
SGP
SGP
THA
THA
THA
THA
AUT
AUT
CHE
CHE
HUN
HUN
Importing
country
MYS
PHL
SGP
THA
IDN
PHL
SGP
THA
IDN
MYS
SGP
THA
IDN
MYS
PHL
THA
IDN
MYS
PHL
SGP
HUN
POL
HUN
POL
AUT
CHE
Year of Trade
Agreement*
1992
1992
1992
1992
1992
1992
1992
1992
1992
1992
1992
1992
1992
1992
1992
1992
1992
1992
1992
1992
1993
1993
1993
1993
1993
1993
28
Country Pairs
Exporting
country
53
HUN
54
HUN
55
HUN
56
NOR
57
NOR
58
POL
59
POL
60
POL
61
POL
62
POL
63
SWE
64
SWE
65
BEL
66
BEL
67
CAN
68
DEU
69
DEU
70
ESP
71
ESP
72
FRA
73
FRA
74
HUN
75
HUN
76
HUN
77
HUN
78
HUN
*Entry into force
Importing
country
NOR
POL
SWE
HUN
POL
AUT
CHE
HUN
NOR
SWE
HUN
POL
HUN
POL
MEX
HUN
POL
HUN
POL
HUN
POL
BEL
DEU
ESP
FRA
IRL
Year of Trade
Agreement*
1993
1993
1993
1993
1993
1993
1993
1993
1993
1993
1993
1993
1994
1994
1994
1994
1994
1994
1994
1994
1994
1994
1994
1994
1994
1994
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
Exporting
country
HUN
HUN
HUN
IRL
IRL
ITA
ITA
MEX
MEX
NLD
NLD
POL
POL
POL
POL
POL
POL
POL
POL
PRT
PRT
USA
COL
COL
MEX
PER
Importing
country
ITA
NLD
PRT
HUN
POL
HUN
POL
CAN
USA
HUN
POL
BEL
DEU
ESP
FRA
IRL
ITA
NLD
PRT
HUN
POL
MEX
MEX
PER
COL
COL
Year of Trade
Agreement*
1994
1994
1994
1994
1994
1994
1994
1994
1994
1994
1994
1994
1994
1994
1994
1994
1994
1994
1994
1994
1994
1994
1995
1995
1995
1995
29